The Investor Dilemma: Rely on Gut or Data to Select the Most Promising Ideas to Invest In
This guest post brings an intriguing perspective to the art of startup-investing by highlighting the role AI and LLMs could play in the selection of the best ideas and teams.
This article expands on the concepts introduced in the September 2023 Otonomist Funding special issue, which presented a transformative approach to equity fundraising.
The approach focused on an online platform that was fully automated, legally binding, transparent, and secure, with blockchain-enabled features and a crypto-first approach.
This article dives into the step preceding equity fundraising and explores how online platforms can be utilized to identify the best projects for investment as an example of finding the best agents in the given ecosystem.
It proposes an approach that combines network theory to analyze the ecosystem dynamics and modern portfolio theory to optimize resource allocation among the most suitable agents.
While the use case for this discussion revolves around a portfolio of early-stage projects seeking venture capital, the concepts presented can be applied to other ecosystems where data is predominantly available online, such as gaming guilds, decentralized autonomous organizations (DAOs), and even traditional corporations.
Conventional agent selection
Throughout history, the key to business success has been delegating the work to the right people. However, finding the best individuals within large ecosystems has always been a challenge. Which ventures will become the next unicorns? Which workers generate the most value?
Traditionally, identifying the top agents in a specific ecosystem has been based on sentiment, intuition, and trust, often relying on personal acquaintances. Over time, decision-making leads to gaining experience, which can then be translated into policies or other types of guidelines.
Venture capitalists still heavily rely on sentiment when selecting projects to invest in. This often leads, as Otonomist reported, to numerous rejections for founders, particularly those without personal connections. Consequently, founders are compelled to dedicate a significant amount of time and energy to networking in order to raise funding for product development, rather than focusing on developing the product itself. To secure funding, simply doing a good job is not enough. But can it be?
An increasing number of activities and interactions are moving to online platforms such as social media (e.g. LinkedIn, Twitter), forums (e.g. Reddit), repositories (e.g. GitHub, Medium), business communication tools (e.g. Google Workspace, Microsoft Office, Zoom, Slack, Telegram, Discord), transactions (e.g. FinTech or Web3), etc.
Having all of the activity and interaction data documented and available online is an opportunity for data-driven analysis to learn about the agents, what they do, how they are perceived, as well as predict the likelihood of them achieving your goals - such as that early-stage venture you invested in becoming a unicorn.
Emerging agent selection
Data from online platforms is diverse and exists in complex structures and relationships. In the past, it has been challenging to comprehensively handle this complexity.
However, recent technological advancements, such as high-quality sentiment analysis using Language Models (LLMs), and cutting-edge discoveries in Machine Learning (ML) on relational databases, as well as network theory, particularly graph neural networks, provide powerful tools for analyzing online content and community dynamics.
These technologies allow us to analyze ecosystems in ways that were previously impossible. Now, we can not only understand the characteristics of the members, but also their reputation, significance, influence, and perception in relation to others.
Additionally, we can measure performance and, in the end, combine all of this information to identify the most qualified or promising agents to achieve our objectives.
However, to accomplish this, we need to understand the ecosystem thoroughly, as well as determine what we want to achieve in the first place. This requires a few paradigm shifts.
Considering ecosystems as dynamic networks. In this representation, each agent represents a node, and their connections and interactions form the links. By adopting this perspective, we can utilize various centrality measures and ML algorithms to identify community dynamics. This analysis helps to understand the importance of projects within the ecosystem, their connections to other projects, and the impact of changes in one project on the overall ecosystem. Similarly, it can be used to evaluate the team members involved in the projects.
Determining the objectives either quantitatively (e.g., market size, acquisition, traction, revenue growth, etc.) or in natural language (e.g., unique selling point, technology used, team potential).
Using the objectives as metrics to identify the best agents, and determine optimal allocation among them. For venture capitalists, this involves identifying the most promising projects and determining the appropriate level of investment for each. With this information, the investor could then proceed with the online equity fundraising process outlined in the September issue of Otonomist.
In broader terms, this approach helps to identify the most suitable agents for the task the principal wants to delegate and determine how to allocate resources among them.
By combining this agent selection approach with performance measurement, a feedback loop can be created to train the ML model and improve its predictions in the future.
Where to start
While this article primarily focuses on identifying the most promising ventures, there are also other use cases to consider.
Efficient agent selection is especially important in decentralized marketplaces that, due to global participation, face the biggest challenges in selecting the best people to carry on tasks and drive business growth.
In fact, Web3 communities that operate primarily or exclusively online are particularly well-positioned to integrate diverse data sources of their members’ activities and employ data-driven approaches to analyze the behaviors and connections within their ecosystem.
This concept is also applicable to traditional large organizations that face challenges in finding or selecting the most suitable teams or workers for specific tasks from their diverse pool of employees.
The non-human agents to choose from, in addition to early-stage projects, include companies (e.g., in portfolio management), information (e.g., in knowledge management), algorithms (e.g., in technology management), or any other context where ML-powered insights can provide data-driven guidance to decision-makers, enabling them to make more informed choices.
While the benefits of identifying the best early-stage projects are straightforward, finding the most suitable people to delegate work to brings multiple layers of advantages.
From a top-down perspective, it ensures a more effective and timely completion of tasks, leading to increased productivity, reduced waste from underutilized talent and, in turn, maximized impact of available resources. Consider a PM in a multinational firm who can simply define the required skills for the project and find the most suitable individuals to involve from the entire organization based on their skills, experience, network and influence.
From a bottom-up perspective, agents are motivated to actively engage and continuously showcase their strengths to become visible for the leadership. By sharing skills and expertise, they can be assigned tasks they excel at and thus unlock their full potential. This approach is particularly significant for decentralized and trustless organizations in Web3.
Ultimately, delegating tasks based on data rather than sentiment, in an objective, transparent, and auditable manner, may lead to a more democratic yet productive ecosystem, resulting in more prosperous outcomes. It will be interesting to observe how this evolution unfolds.
By Zarja Hude (©2024. All rights reserved.) Zarja works in Ernst & Young’s LegalTech and Innovation group in London after graduating from Cambridge University in Corporate Law. She is an active member of the Cambridge Blockchain Society and organizes its Weekly Brunches, a Saturday workgroup co-hosted with King's College Entrepreneurship Lab aimed at helping students bring their innovative ideas to life.
> Join the Otonomos official “Towards Full Otonomy” Telegram group to compare notes with other users and our wider community!